MNRAS 000, 1–11 (2021) Preprint 14 June 2021 Compiled using MNRAS LATEX style file v3.

Comparing Compact Object Distributions from Mass- and Presupernova

Core Structure-based Prescriptions
Rachel A. Patton 1★ , Tuguldur Sukhbold1,2 †, and J.J. Eldridge3
1 Department of Astronomy, The Ohio State University, 140 West 18th Ave, Columbus, OH 43210, USA
2 Centerfor Cosmology and AstroParticle Physics, The Ohio State University, 191 West Woodruff Avenue, Columbus, OH 43210
3 Department of Physics, University of Auckland, Private Bag 92019, Auckland, New Zealand
arXiv:2106.05978v1 [astro-ph.HE] 10 Jun 2021

Accepted XXX. Received YYY; in original form ZZZ

ABSTRACT
Binary population synthesis (BPS) employs prescriptions to predict final fates, explosion or implosion, and remnant masses
based on one or two stellar parameters at the evolutionary cutoff imposed by the code, usually at or near central carbon ignition.
In doing this, BPS disregards the integral role late-stage evolution plays in determining the final fate, remnant type, and remnant
mass within the neutrino-driven explosion paradigm. To highlight differences between a popular prescription which relies only
on the core and final stellar mass and emerging methods which rely on a star’s presupernova core structure, we generate a
series of compact object distributions using three different methods for a sample population of single and binary stars computed
in BPASS. The first method estimates remnant mass based on a star’s carbon-oxygen (CO) core mass and final total mass.
The second method uses the presupernova core structure based on the bare CO-core models of Patton & Sukhbold (2020)
combined with a parameterized explosion criterion to first determine final fate and remnant type, then remnant mass. The third
method associates presupernova helium-core masses with remnant masses determined from public explosion models which rely
implicitly on core structure. We find that the core-/final mass-based prescription favors lower mass remnants, including a large
population of mass gap black holes, and predicts neutron star masses which span a wide range, whereas the structure-based
prescriptions favor slightly higher mass remnants, mass gap black holes only as low as 3.5 M  , and predict neutron star mass
distributions which cluster in a narrow range.
Key words: stars: evolution – stars: massive – stars: neutron –stars: black hole – supernovae: general

1 INTRODUCTION uncertainties and limit the computational expense, BPS codes termi-
nate a given massive star’s evolution at or near carbon ignition in the
To understand the properties of how massive stars die and the com-
core. The commonly used “rapid” type of BPS codes do not actively
pact objects they leave behind, we must understand how they live.
evolve their stars at all, relying instead on the semi-analytic approx-
Any prediction of the occurrence rate and formation channels of phe-
imations of Hurley et al. (2000) and Hurley et al. (2002) that only
nomena involving neutron stars (NSs) and stellar mass black holes
approximately track a few global parameters until the asymptotic
(BHs) requires knowledge of the underlying birth distribution of such
giant branch.
compact objects and the relative rates of the events (supernovae or im-
plosions) which produce them. Because nearly all massive (>8 M  )
stars live in binaries or higher order systems (e.g., Sana et al. 2012,
2013; Moe & Di Stefano 2017), the effects on these distributions
caused by interaction with a companion must be considered since it
can complicate a star’s evolution, final fate, and associated compact
object (e.g., De Marco & Izzard 2017; Schneider et al. 2021).
Binary population synthesis (BPS) is often employed to capture
these effects, such as mass transfer, common envelope evolution, and
merging, for large, quasi-realistic populations of stars. High mass
stars are particularly challenging to model because the physics gov-
erning their evolution, especially beyond core carbon ignition, re-
mains uncertain and computationally expensive. To mitigate these
★ E-mail: patton.502@osu.edu
† NASA Hubble Fellow
© 2021 The Authors
Once evolution ceases, a final fate is assumed based on either ini-
tial or final properties of the star. This includes separating explosions
from implosions, thus NSs and BHs, through an arbitrary value of
initial mass (Sravan et al. 2019), helium core mass 𝑀He (Zapartas
et al. 2019), simple energy arguments (Eldridge & Tout 2004b),
and carbon-oxygen (CO) core mass 𝑀CO and final stellar mass
𝑀fin (Fryer et al. 2012, hereafter F12), the most commonly used pre-
scription. The primary shortcoming of these simplistic approaches
is their failure to capture the effects late-stage evolution has on a
star’s core structure, and ultimately, its demise. An extensive body
of work shows that a star’s final fate, within the neutrino-powered
explosion paradigm, depends sensitively on the structure of a star’s
core immediately preceding the iron core–collapse, which in turn is
set by the last few thousand years of evolution in the stellar core (e.g.,
Woosley et al. 2002; O’Connor & Ott 2011; Ugliano et al. 2012; Pe-
jcha & Thompson 2015; Sukhbold et al. 2016; Sukhbold & Adams
2 Patton et al.
2020). Neglecting these stages leads to final fate and compact object
predictions based on properties that may not strongly correlate with
a star’s actual explodability.
Recently Patton & Sukhbold (2020, (hereafter PS20) proposed a
method to base a BPS prescription for final fates on a star’s pre-
collapse core structure. The structure is based on bare CO-cores,
whose evolutionary tracks are set by their starting mass and compo-
sition, evolved from central carbon ignition to core–collapse, filling
in the evolutionary stages missed by the BPS code. The user can
then infer a final structure for any star given their 𝑀CO and its start-
ing composition, and apply their preferred structure-based criteria
to reflect modern results on neutrino-driven explosion simulations.
Since the PS20 approach requires knowledge of the initial CO-core
composition, something which is not tracked in rapid BPS codes,
currently it is only applicable to BPS calculations that are based on
full stellar models.
However, there is an emerging alternative method, which we refer
to as “hybrid”, that applies results from detailed neutrino-driven ex-
plosion calculations to BPS simply through the terminal 𝑀He . This
mass transfer, common envelope evolution, mergers, orbital decay,
and ultimately, the final fates of massive stars, and its underlying
stellar models are based on those of Eggleton (1971) and Pols et al.
(1995).
Version 2.2 is largely equivalent to version 2.1 (Eldridge et al.
2008; Eldridge & Stanway 2009; Eldridge et al. 2017) but has greater
mass resolution between single star models than before. Binary mod-
els have coarser mass resolution than the single models. More impor-
tantly, BPASS v2.2 incorporates the updated binary birth parameter
distributions of Moe & Di Stefano (2017, see their Table 13). BPS
codes typically assume that a binary’s initial conditions, the mass
of the primary (𝑀1 ), the mass ratio (𝑞 = 𝑀1 /𝑀2 ) between the two
components, and the orbital period (𝑃), are independent, with 𝑞 and
𝑃 having flat distributions and primary mass being drawn from an
assumed initial mass function (IMF). However, there are strong ob-
served trends in both 𝑞 and 𝑃 with respect to primary mass, which
BPASS now considers.
A detailed description of the initial model grid is given in Stanway
& Eldridge (2018). Our set of single BPASS models span 0.2-100
approach utilizes the correlation pointed out by Ertl et al. (2020) M  in steps of 0.1 M  and our binary models span 4.5 – 100 M  for
between the presupernova 𝑀He and the properties of core–collapse,
which allows BPS codes to determine the final properties based
implicitly on the presupernova core structure without actually com-
puting this structure (unlike in PS20). Recently Woosley et al. (2020)
compiled a table directly connecting 𝑀He to the compact remnant
masses based on the calibrated neutrino-driven explosion models
of Ertl et al. (2020) and pulsational pair-instability models from
Woosley (2019), and used it to infer the birth mass distributions of
NSs and BHs in various simple populations. Román-Garza et al.
(2020) adopted a similar method, instead compiling the table based
on the results from Sukhbold et al. (2016), to predict BH-NS merger
rates from a detailed population of full binary stars and comparing
them to those predicted from F12. Since the presupernova 𝑀He is
well determined by the time of central carbon ignition, this simple
and efficient method can be directly applied in any rapid BPS model.
To understand how the results based on these different prescrip-
𝑀1 with increments of 0.5 M  . For the binaries, each primary star
cycles through each possible combination of mass ratio and orbital
period, with 𝑞 ranging between 0.1 and 0.9 in increments of 0.1, and
log 𝑃 ranging between 1.4 and 4 days in steps of 0.2 dex. We ignore
binary models with log P below 1.4 to eliminate all merging systems,
and to build a uniform coverage over the mass space we interpolate
between models for each combination of 𝑞 and log 𝑃. The twin star
systems (𝑞 = 1) in BPASS are handled by incorporating its statistics
from Moe & Di Stefano (2017) into the systems with 𝑞 = 0.9. All
models are at solar (𝑍 = 0.02) metallicity. Stars are evolved until
core carbon ignition, where we extract the total mass of the star 𝑀fin ,
helium and CO- core masses 𝑀He , 𝑀CO , carbon mass fraction 𝑋C ,
which we adopt as our proxy for composition, and the number of
stars per million solar masses with a given set of initial conditions,
as determined by an IMF and the Moe & Di Stefano (2017) weights.
BPASS assumes a Kroupa (2001) IMF for masses between 0.5 and
tions vary from one another it is important to apply them to the 1 M  , and a Salpeter (1955) IMF for stars with masses heavier than
same sample population of stars. In this study, we apply the F12,
PS20, and hybrid prescriptions to a suite of BPASS (Eldridge et al.
2008; Eldridge & Stanway 2009; Eldridge et al. 2017; Stanway &
Eldridge 2018) single and binary models, comparing the distribution
of compact objects determined from each prescription. In §2 we dis-
cuss the BPASS models, the differences between the F12 and PS20
prescriptions, and how we construct the distributions. §3 compares
the F12 and PS20 distributions. In §4 we examine how the hybrid
and PS20 distributions compare and we address the impact of the
differences between the three prescriptions in §5. Finally, our results
are summarized in §6.
2 MODELS
2.1 BPASS models
We construct our distributions from a set of BPASS (v2.2) models,
1 M.
A few minor corrections were manually applied to the BPASS grid
before the interpolation. Six single star models had anomalously
low values of 𝑀CO and/or 𝑋C . We believe these were errors with
the models and replaced them with 𝑀CO and 𝑋C values linearly
interpolated over the set of single star models. There were 25 entries
missing from the binary models and an additional two binary models
with anomalously low 𝑋C , again, numerical errors with those specific
models. Each missing or spurious model was replaced by an average
of the models of the same initial orbital period and mass with mass
ratio 0.1 higher and lower than the missing value. If the missing
model had 𝑞 = 0.1, we averaged the model with 𝑞 = 0.2 of the same
mass and orbital period with the 𝑞 = 0.1 model of the same mass
with the next lowest orbital period. If the missing model had 𝑞 = 0.9,
we averaged the model with 𝑞 = 0.8 of the same mass and orbital
period with the 𝑞 = 0.9 models of the same mass and next highest
orbital period.
Fig. 1 highlights key properties of these models, comparing 𝑀fin ,
106 M  of single stars and 106 M  of binary systems. BPASS is a
hybrid binary population synthesis code which evolves its single stars
and the primaries of binary stars with a detailed stellar evolution code
while approximating the evolution of the secondaries using the semi-
analytic evolution equations from Hurley et al. (2000, 2002), then
evolving the secondaries in detail once the primary star’s evolution
terminates. BPASS implements prescriptions for handling winds,
MNRAS 000, 1–11 (2021)
𝑀He , and 𝑀CO as a function of initial mass (𝑀init for single stars, 𝑀1
for binaries) for the single stars in the top panel and sample subsets
of binary stars, at fixed mass ratios and initial periods, in the bottom
panel. All binary models experience some amount of mass loss via
transfer onto a companion, leading to the final mass and He-core mass
to either be much closer than in the single stars, or identical. After the
turnover, the final mass converges with helium core mass and marks
 Compact Object Distribution Comparison 3

16 of stars have their core structures affected by binary interaction and

wind mass loss.
14
Fig. 2 shows where the single and binary stars fall in this plane.
12 There are two main sequences, one spanning between 1.4 - 14 M 
in 𝑀CO and 0.15 - 0.3 in 𝑋C that forms an upper bound, and the
Mass (M )

10
other ranging between 1.4 - 9.5 M  and 𝑋C of 0.10 - 0.25 bounding
8 the points from the bottom. The lower curve, which we label the
“normal" sequence, is where stars fall if they retain some or all
6
of their envelope. The upper curve, which we label the “stripped"
4 sequence, corresponds to stars which have lost all of their envelope,
MHe
either by wind mass loss or mass transfer onto a binary companion.
2 Mfin
The characteristic curve shape is set by a few key aspects of stellar
MCO
0 physics. In our set, the core mass generally increases with initial
0 20 40 60 80 100 mass, and becomes less degenerate. The higher the core mass, the
Minit (M ) lower the central density, and the effect of C12 (𝛼, 𝛾) O16 is stronger
during core helium burning, leading to decreased 𝑋C . For more
25
degenerate lower mass cores, carbon burning via the 3𝛼 process is
strong, increasing 𝑋C . This is true whether or not the envelope has
been stripped. However, the helium core ceases to grow when the
20 q = 0.5 logP = 2
envelope is completely stripped. Instead it recedes, leaving behind a
(+ 10 M )
carbon gradient and resulting in less helium being brought into the
Mass (M )

15
core. 𝑋C increases, causing the offset between the two sequences.
Between the two sequences, we see turnovers in 𝑋C . Each turnover
roughly corresponds to a subset in our population, with turnover
10
mass point corresponding to its lowest mass star which has its enve-
lope entirely stripped. For the single stars, this is the 27 M  model
5
MHe (𝑀CO ≈ 9M  ). With the binaries, envelope stripping can occur at
q = 0.6 logP = 3 Mfin much lower primary masses depending on the star’s initial orbital
MCO period. The scattered binary points between the two sequences cor-
0
0 20 40 60 80 100
respond to the turnover populations of binaries, e.g., stars with little
M1 (M ) or to envelope at the same 𝑞 and log 𝑃, but varying 𝑀1 . A shorter
initial period corresponds to a lower mass turnover. Each of these
Figure 1. A comparison of the helium core (blue), CO-core (green), and populations follow roughly the same trends as the single stars. Mod-
final mass (red) of stars for the single models (top) and a subset of the binary els of increasing initial mass will fall on the normal sequence until a
models (bottom). The q = 0.5, log P = 2 models have been offset by +10 M to model has its envelope entirely stripped. Then stars of higher initial
have no overlap with the other models. The convergence of 𝑀fin and 𝑀He
mass will turnover onto the stripped sequence.
at ∼27 M for single stars is caused by envelope stripping. Above this mass,
all stars have lost their envelopes to winds and will ultimately experience
collapse as bare cores. In binary models, the minimum mass at which a star
loses its envelope depends heavily on the initial orbital period.
the minimum birth mass for complete envelope stripping. For single
Mergers can complicate the picture a little more. Though merged
binary systems were removed from the sample population, an initial
analysis revealed that they tended to fall in the exact same place as the
single stars. We suspect this is due to when the stars merge. If stars
merge on the main sequence, we expect them to evolve equivalently
to single stars of their new mass, meaning they will lie on the normal
sequence until wind stripping causes a turnover, where higher mass
stars this happens at around 27 M  , and depending on the initial models fall on the stripped sequence. However, there was a separate
period it ranges between 4.5 M  (log 𝑃 = 1.4) and 30 M  (log 𝑃 =
4). After this convergence, 𝑀fin , 𝑀He , and 𝑀CO generally increase
with initial mass, but non-monotonically due to the adopted wind
prescriptions. BPASS uses the Wolf-Rayet mass loss prescriptions
of Nugis & Lamers (2000), which vary between WC and WN stars.
group of models with values of 𝑋C which fell below the normal
sequence, all of which were merger products. We suspect these are
models which merged post-main sequence but cannot verify this
since we only know the population’s initial and final conditions. If
these are post-main sequence mergers, it is worth investigating their
The second mass peak around 80 M  is due to significant mass loss
experienced by stars of higher mass on the main sequence. These
stars bypass the red supergiant phase, instead evolving straight to
Wolf-Rayet stars after the completion of core hydrogen burning.
Because BPASS evolves the full structure of stars in detail, we
are able to keep track of 𝑋C in addition to 𝑀CO , both of which
are measured at carbon ignition when evolution terminates. PS20
highlighted that the evolutionary history of a star writes itself into
the core by changing the CO-core mass and its starting composition.
It has also been shown that stars tend to fall in well-defined curves in
𝑋C - 𝑀CO space set by the physics of that evolutionary history (e.g.,
PS20, Schneider et al. 2021; Laplace et al. 2021). By characterizing
impact on core structure. A preferential decrease in 𝑋C results in
cores more difficult to explode and could impact the relative rates of
supernovae and BH formation.
2.2 Predicting remnant masses and building the distributions
While BPASS is capable of evolving stars to neon ignition, prop-
erties of the stars were extracted at carbon ignition for a consistent
application of various models considered in this study. We recorded
the initial mass, final mass, helium core mass, CO-core mass, com-
position of the CO core, and the star’s IMF weight, and applied
them to estimate compact remnant masses through various prescrip-
our models in this plane, we show for the first time how populations tions. We consider 1.4 M  to be the minimum 𝑀CO to experience

MNRAS 000, 1–11 (2021)

4 Patton et al.
4.0
the two-parameter criterion of Ertl et al. (2016, 2020) on the predicted
0.30 Binary stars final CO-core structures. This entails measuring 𝑀4 , the enclosed
Single stars mass at the place where specific entropy equals 4, and 𝜇4 , a radial
3.5
gradient in mass at the point of 𝑀4 , in the presupernova profile and
0.25
Stripped sequence
comparing them to their W18-engine based results (using 𝑘 1 = 0.182
3.0 and 𝑘 2 = 0.0608). For stars that were deemed to produce a supernova
we take the 𝑀4 value as the baryonic NS mass, and for stars that

log P
0.20
XC

2.5
failed to explode we adopt the helium core mass value as BH mass.
Scenarios where the star explodes but leaves behind a BH were not
0.15 explored. See section 3.2 of PS20 for further details on this sample
Normal sequence Turnover 2.0 implementation.
The PS20 models span 2.5 to 10 M  in 𝑀CO , a narrower range than
0.10
the BPASS models. Since a successful neutrino-powered explosion
1.5
above 𝑀CO > 10 M  is unlikely, we simply assume they all implode
2 4 6 8 10 12 14
to produce a BH, and adopt 𝑀He for its mass. For stars with 𝑀CO
MCO (M )
between 1.4 and 2.5 M  , all stars are assumed to explode and produce
0.30 a NS no matter the starting composition of the CO-core. For these
0.8 stars the baryonic NS mass is determined by directly interpolating
in 𝑀CO over the results of the lightest models from Sukhbold et al.
0.25 fractional remaining envelope (2016).
0.6 A second set of distributions were computed through the delayed,
binary prescription laid out in section 4.3 of F12. This method esti-
0.20
XC

mates the remnant mass based on the CO-core mass and final mass of
0.4
the star. The prescription first estimates a remnant’s initial baryonic
0.15 mass based on a star’s CO-core mass. Then additional mass is added
to the remnant, accounting for fallback, based on both the CO-core
0.2
mass and final mass of the star. The distinction between NS and BH is
0.10 left to the user, as the prescription solely estimates mass of a compact
remnant. We assign a cutoff in baryonic mass of 2.5 M  (≈ 2.3 M  in
0.0
2 4 6 8 10 12 14 gravitational mass), with NSs having masses less than the cutoff and
MCO (M )
Figure 2. 𝑋C and 𝑀CO values for the populations of single and binary stars
from BPASS. Top: Binary models are color coded by their initial period, and
all single stars are highlighted in black. Bottom: All single and binary models
are color coded by the amount of hydrogen envelop mass retained at the
time of carbon ignition. The upper bound (stripped sequence) of the models
are made up of stars without any envelope, lost all through mass transfer
onto a binary companion and/or wind mass loss. The lower bound (normal
sequence) corresponds to the expected relationship between 𝑀CO and 𝑋C for
stars whose helium core evolves either in or effectively in isolation. The
BHs having masses above. F12 do suggest baryonic-to-gravitational
mass conversions through Lattimer & Yahil (1989) for NS, and an
arbitrary 10% reduction for BHs, but we refrain from adopting these
methods to remain consistent across prescriptions. We only convert
NS masses through Lattimer & Prakash (2001), and keep the BHs
masses in baryonic form.
Fig. 3 shows how the predicted remnant masses compare between
F12 and PS20 across the 𝑀CO range of our BPASS models. The left
hand panel compares the remnants produced by single stars and the
right hand panel compares the remnants from binary systems. There
upward streams (turnover) connecting the two sequences, at around 10 M is very little overlap between the two. The F12 NS masses increase
for single stars and at lower values for binaries depending on the initial period, with increasing 𝑀CO , ranging from 1.2 to 2.2 M  , whereas the PS20
marks the transition from regular to stripped sequence.
core–collapse (e.g., Woosley & Heger 2015). For simplicity, we have
completely ignored the effects of natal kicks, and our compact rem-
nant distributions directly describe the final fate for each star in our
population. Unless the NS mass is directly determined in gravita-
tional form, we convert all baryonic NS masses into gravitational
masses using the methods in Lattimer & Prakash (2001) with an
NS masses mostly cluster between 1.2 and 1.6 M  . In the binary
panel, we see an additional group of higher mass NSs predicted by
PS20, created in an island of explodability at higher birth mass (e.g.,
Sukhbold & Woosley 2014; Sukhbold et al. 2018). In general, NS
masses in F12 have a nearly unique correspondence with 𝑀CO , while
in PS20 similar remnants can be made by stars that are far apart in
birth mass, and conversely, stars of similar birth masses can result in
different remnants as well.
The lowest mass BHs vary between prescriptions, with F12 pre-
adopted NS radius of 12 km. BH masses remain baryonic, though dicting BHs down to 2.5 M  for both the binary and single mod-
the difference between baryonic and gravitational mass is small. All els, where as the PS20 prescription predicts BHs down to 3.5 M  ,
distributions are normalized to 1, and NS masses are grouped in bins the helium-core mass of the lightest imploding model. The highest
of 0.1 M  and BH masses are grouped in bins of 0.5 M  . mass BHs agree between both prescriptions. Additionally, there is a
To create the PS20 based distributions, we first took the 𝑀CO
and 𝑋C values for each star to obtain their presupernova structures.
𝑀CO was measured at the place where the helium mass fraction
dropped to 10%. The CO cores of BPASS stars are not pure mixes of
carbon and oxygen alone, like the idealized models from PS20 are,
but the contribution to its initial composition from other elements is
insignificant. The final fates for each star are determined by applying
MNRAS 000, 1–11 (2021)
turnover in remnant mass around 𝑀CO = 9.5 M  , caused by envelope
stripping, apparent in both prescriptions, but it is more pronounced
in the F12 remnants. This turnover marks the transition between
the two sequences shown in Fig. 2, with every remnant above the
turnover coming from a stripped star. The BH masses from F12 also
increase more quickly with 𝑀CO than in the PS20 models due to the
dependence of the remnant mass on 𝑀fin , which is more strongly af-
 Compact Object Distribution Comparison 5
Single Binary Single Binary
15.0 PS20 15.0 15.0 PS20 15.0

F12 Hybrid
12.5 12.5 12.5 12.5
Mbh

Mbh
10.0 10.0 10.0 10.0

7.5 7.5 7.5 7.5

5.0 5.0 5.0 5.0

2.5 2.5 2.5 2.5

2.2 2.2 2.2 2.2

2.0 2.0 2.0 2.0

Mns, grav

Mns, grav
1.8 1.8 1.8 1.8

1.6 1.6 1.6 1.6

1.4 1.4 1.4 1.4

1.2 1.2 1.2 1.2

1.0 1.0 1.0 1.0

5 10 15 5 10 15 5 10 15 5 10 15
MCO (M) MCO (M)
Figure 3. Remnant masses determined by the F12 (gold triangles) and PS20
(blue circles) prescriptions as a function of 𝑀CO for each BPASS model,
single (left) and binary (right), capable of reaching core–collapse. NS grav-
itational masses are shown on the bottom, and baryonic BH masses on the
top; note the vertical axis break. The greater diversity in outcomes in the
right panel demonstrates the impact of binary interaction on the core. The
steady increase in remnant mass indicates that the F12 prescription effec-
tively functions as a cut in 𝑀CO , separating the NSs (explosions) from the
BHs (implosions).
fected by mass loss from winds or binary interaction than is 𝑀He . BH
masses only agree for the highest mass stars because their envelopes
have been completely stripped; 𝑀fin = 𝑀He . In addition, more appar-
ent in the binary panel, the PS20 models predict a population of NSs,
due its consideration of core structure, for stars that were deemed to
produce a BH according to F12.
These discrepancies point to the key issue with the F12 method
and any prescription that ignores late-stage evolution: the failure to
consider the role presupernova core structure plays in determining
final fate and remnant type (e.g., Woosley et al. 2002; O’Connor & Ott
2011; Ugliano et al. 2012; Pejcha & Thompson 2015; Sukhbold et al.
2016; Sukhbold & Adams 2020). That presupernova core structure,
in turn, depends on how the CO-core evolves from carbon ignition
to collapse, and the evolutionary path taken by the CO-core depends
dominantly on its mass and initial composition. But because BPS
codes often terminate at carbon ignition, the stages of evolution
most important in determining final core structure are precisely the
stages of evolution disregarded in the models. Prescriptions which
only predict final fate and remnant properties based on a star’s mass
disregard the critical impact on structure by the CO-core’s starting
composition. Two stars with identical CO-core masses but different
core compositions could have dramatically different outcomes.
Finally, we apply the hybrid prescription. The benefit of this
method is its compatibility with every BPS code, unlike the PS20
models, which only work with codes that employ active stellar evo-
lution. We construct a similar table as in Román-Garza et al. (2020),
by taking the pre-supernova helium-core masses from Sukhbold et
al. (2016) and their associated baryonic NS masses, which include
MCO (M) MCO (M)
Figure 4. A comparison of the predicted remnant masses as a function of
CO-core mass for the PS20 (blue circles) and hybrid (orange stars) prescrip-
tions. The left panel shows the results from the single star models and the
right panel, the binary models. The two prescriptions predict nearly identical
BH masses but vary somewhat in NS mass predictions. The maximum 𝑀CO
which produces a NS depends on the underlying models’ islands of explod-
ability, which depend on the adopted evolutionary physics. NS masses are
gravitational, while BH masses are baryonic; note the axis break.
was adopted. If that remnant was a NS, then we adopt the NS mass
from the constructed table. If that model resulted in an implosion, we
adopt 𝑀He of the BPASS models as the remnant mass. Because the
tabulated remnant masses already reflect outcome, explosion versus
implosion, we do not need to impose a cutoff between the two.
Fig. 4 compares the predicted remnant masses for both the PS20
and hybrid prescriptions. The slight discrepancy in NS masses from
stars with 𝑀CO below 2.5 is due to interpolating over the remnant
masses from (Sukhbold et al. 2016) using 𝑀CO (for PS20) and 𝑀He
(for hybrid). Because both methods adopt 𝑀He as the BH mass, the
agreement between the two methods for BHs is very good. There
are only slight discrepancies, where the hybrid method predicts a
small island of explodability for the single star population at higher
𝑀CO , and the PS20 method does not. The location and extent of such
islands depends on the models the user selects to build the table; see
Sukhbold et al. (2018) for a discussion on the differences between
underlying models used in PS20 and hybrid prescriptions.
3 COMPARING THE PS20 AND F12 DISTRIBUTIONS
Differences between the two distributions are significant for both the
single and binary populations. In this section we highlight the key
properties between the two.
3.1 Single star distributions
any late time fallback. For models below 12 M  we adopt the Z9.6
explosion engine, and above we use the W18 engine. Non-exploding
models were assigned 𝑀He as their remnant mass. We then inter-
polated over the table using 𝑀He from both the single and binary
star BPASS models. If a helium core mass fell between two models
with different outcomes, the remnant type of the lower bound model
Previous works employing combination of detailed stellar evolution
calculations until iron core–collapse and calibrated neutrino-driven
explosion models have shown that remnant distributions from sin-
gle stars qualitatively match observed NS and BH mass functions
(Sukhbold et al. 2016, 2018; Ertl et al. 2020). However, these dis-
tributions have not been compared to those predicted from common
MNRAS 000, 1–11 (2021)
6 Patton et al.

101 weighted average mass of 1.39 M  ), comparable to the observed

PS20 peak of the NS mass function (e.g., Özel et al. 2012; Özel & Freire
F12 2016) and to the predictions based on prior detailed calculations
(e.g., Timmes et al. 1996; Pejcha & Thompson 2015). In contrast,
100
the F12 distribution has very different shape and predicts a substan-
tially smaller value for the most common NS. The peak is located at
1.2 M  , the smallest value computed in the prescription, and sharply
dM
dN

drops off with increasing NS mass, though the average NS mass is

10−1
1.35 M  . We expect this behavior because NS masses from the F12
prescription have a nearly monotonic dependence on 𝑀CO (Fig. 3),
which generally increases with initial mass (Fig. 1) while 𝑑𝑁/𝑑𝑀 de-
creases, resulting in more low-mass NSs. Since the F12 prescription
only depends on 𝑀CO and 𝑀fin it is most sensitive to the properties
10−2
1.0 1.2 1.4 1.6 1.8 2.0 2.2 of the underlying stellar population. While the PS20 prescription is
Mns, grav also sensitive to the underlying population, its dependence on core
structure mitigates some of that sensitivity because stars with differ-
PS20 ent initial masses can have similar core structure, and thus similar
F12 NS masses. There is also a gap in the F12 distribution caused by the
transition between two different fallback mass calculations. The NS
stars with masses above the gap have their fallback masses calculated
10−1
depending on both 𝑀fin and 𝑀CO , whereas the fallback mass for NSs
dM
dN

below the gap depends on CO-core mass alone.

3.1.2 Black hole distribution

10−2
There is more qualitative agreement between the F12 and PS20 pre-
scriptions in the BH distributions than in the NS distributions. In
4 6 8 10 12 14 16
Mbh accord with the predictions from Sukhbold et al. (2016), both dis-
tributions shown in the middle panel of Fig. 5 peak around ∼10
PS20 with Mfin M  , though the IMF weighted average remnant masses are 8.59 and
F12 7.52 M  for the PS20 and F12 distributions respectively. Both dis-
tributions also nearly match in predicted frequencies of BHs above
12.5 M  . The progenitors of these systems have their envelopes en-
10−1 tirely stripped by winds, causing the final stellar mass and 𝑀He to be
equivalent (stars above 27 M  , see top panel of Fig. 1).
dM
dN

But below 12.5 M  the two distributions are most discrepant

because of the differences in how BH mass is determined. The PS20
distribution is only sensitive to the underlying population’s He-cores
10−2
since 𝑀He is adopted as 𝑀bh . It experiences a drop in frequency
starting at 11.5 M  , the local maximum mass at which the He-cores
4 6 8 10 12 14 16
of the single star population reach before turning over. The He-cores
Mbh do not grow again until they shrink to 9 M  , coming from stars with
initial masses above 43 M  . But these stars are so massive, the IMF
weighting strongly disfavors them, meaning an overall decrease BH
Figure 5. Compact object distributions for a population of 106 M of single frequency for BHs above 11.5 M  .
BPASS models determined by the PS20 (blue) and F12 (gold) prescriptions.
The F12 models are also sensitive to these effects because, al-
The top panel shows the NS distribution. The center panel shows the black
though the F12 distribution determines BH mass from 𝑀fin , it is
hole distribution. The bottom panel again shows the black hole distribution
with the PS20 prescription adopting 𝑀fin instead of 𝑀He as the black hole equivalent to helium core mass (𝑀fin = 𝑀He ) for stars with birth
mass. The intrinsically different shapes of the NS distributions point to the masses above 27 M  . However, 𝑀fin turns over before 𝑀He does. At
non-negligible differences in how each prescription determines its NS mass. 21 M  the single stars reach a global maximum in 𝑀fin , then drop
off (Fig. 1). Decreasing 𝑀fin decreases 𝑀bh , leaving a deficit of BH
prescriptions employed in BPS codes. We present below a brief com-
parison of the F12 and PS20 predicted compact remnant distributions
based on single star population.
3.1.1 Neutron star distribution
masses down to 10 M  instead of 11.5 M  in the PS20 distribution.
Lower-mass BHs also exhibit an important difference between
the two distributions, namely the prediction of lower-mass mass gap
remnants. The simple implementation of PS20 in this study ignores
all fallback (remnant mass is either 𝑀4 or 𝑀He depending on the
outcome), the smallest BH mass is effectively the smallest Mhe that
implodes (3.5 M  ). However, in F12 the remnant mass continuously
The top panel of Fig. 5 shows the distribution of gravitational NS increases with Mco, and the distributions predict a substantial pop-
masses predicted by the PS20 and F12 prescriptions, revealing two ulation of BHs as low as 2.5 M  , the adopted lower limit. Though
key differences. The PS20 results show a narrow distribution ranging there is some evidence for mass gap BHs, such as those predicted
between 1.2 and 1.6 M  , and exhibit a sharp peak at 1.4 M  (IMF- by Ertl et al. (2020) and the observed BH - red giant binaries from

MNRAS 000, 1–11 (2021)

 Compact Object Distribution Comparison 7
Thompson et al. (2019); Jayasinghe et al. (2021), as well as new
observational techniques to detect them (Jayasinghe et al. 2021), it PS20
remains unclear how common they are and what are their origins. 101 F12
In addition to comparing the distributions from the two prescrip-
tions as is, we also augment a version of PS20 prescription to adopt 100
the final stellar mass instead of 𝑀He as the BH mass, shown in the
bottom panel of Fig. 5. We do this first and foremost because of the 10−1

dM
dN
uncertainty in how much of the total stellar mass ultimately ends up
in the BH. For the loosely bound cool envelopes of red supergiants, 10−2
Lovegrove & Woosley (2013) suggest that neutrino emission could
unbind the envelope, thus 𝑀BH ∼ 𝑀He . However, for the more tightly
10−3
bound compact envelopes of blue supergiants, the neutrino emission
might not be sufficient to eject any part of the envelope. In this in-
stance, 𝑀fin would be a better estimate of the BH mass. Ideally, for 10−4
1.0 1.2 1.4 1.6 1.8 2.0 2.2
a more careful treatment one would need to use the envelope struc- Mns, grav
ture at core–collapse (or carbon ignition, since the envelope does not
have time adapt to the changing core) and consider the binding en- Figure 6. NS distribution from stars originally in binary systems determined
ergy profile in order to adopt 𝑀He or 𝑀fin or somewhere in between by the F12 (gold) and PS20 (blue) prescriptions. Because the F12 prescription
for the BH mass. predicts increasing remnant mass with 𝑀CO and initial mass, the distribution
extends to higher NS mass.
In the absence of this knowledge, we show here how the BH
distribution changes with 𝑀fin adopted as the BH mass. This also
provides a slightly better comparison to the F12 prescription, which
loss or mass transfer onto a companion. The lack of new helium
uses the final mass of the star to determine the amount of fallback
being brought into the core combined with the carbon gradient left
mass on higher mass CO-cores, where the most massive cores (>11
behind by the receding convective core results in higher 𝑋C for a
M  ) produce BH masses equivalent to the final mass of the star and
given 𝑀CO . Having a higher 𝑋C extends the range of 𝑀CO which
stars with 6 < 𝑀CO < 11 M  produce BHs with masses set by some
can produce NSs.
𝑀fin -dependent fraction of the final mass. Stars with 𝑀CO < 6 M 
The F12 prescription is not sensitive to these changes in 𝑋C ,
have either a fixed or 𝑀CO -dependent amount of fallback. This leads
instead reflecting binary interaction only in changes to 𝑀CO and
to a greater discrepancy between the PS20 and F12 distributions
𝑀fin . The binary NS distribution still extends to 2.2 M  and, up
because F12 produces a spectrum of fallback masses whereas PS20
to 1.8 M  it is nearly identical in frequency to the single F12 NS
either adopts all or none of the envelope mass in addition to 𝑀He .
distribution. However, mass loss from binary interaction has shrunk
For the PS20 models, every star which retained some or most of its
core masses for some stars, leading to an over production of higher
envelope shifts upward in mass by about 5 M  , which dramatically
mass NSs (1.8-2.2 M  ), hence the higher frequency in this mass
increases the frequency of high mass BHs. This new distribution is
range. Combinations of 𝑀CO and 𝑀fin capable of producing a NS of
comparable to the equivalent 𝑀BH = 𝑀fin predictions of Sukhbold
a given mass can now come from a broader range of initial masses
et al. (2016).
of their progenitors depending on how much mass a star loses or
gains from binary interaction. The end result is a near continuous
distribution of NSs.
3.2 Binary star distributions
Because binary interaction can substantially change the properties
of the CO-core, we expect to see deviations from the single star 3.2.2 Black hole distribution
distributions, such as higher mass remnants. Below we examine the
The binary distributions are similar to the single star ones, just with
extent to which binary interaction shapes the distributions.
slight reduction in maximum BH mass, 15M  instead of 15.5 M  ,
caused by extreme mass loss. However, the highest mass BHs from
binaries slightly increase in frequency over those from single star
3.2.1 Neutron star distribution
progenitors. This is caused by shallower decrease in 𝑀fin and 𝑀He
The differences between the F12 and PS20 prescriptions are most in the binary models for stars above 80 M  (see Fig. 1). The PS20
apparent in the NS distribution from binary stars. Again, the PS20 models show no strong peak whereas the F12 distribution tends to
distribution peaks near a NS mass of 1.4 M  and retains a more favor lower mass BHs, again predicting high frequency of remnants
normal shape. The F12 distribution is more extended, with a slight down to 2.5 M  . The average remnant masses are similar to the
peak at 1.2 M  . Still, both prescriptions produce average NS masses single star distribution, 8.16 M  for the PS20 distribution and 7.33
of 1.38 M  . M  for the F12.
As shown in Fig. 6, the PS20 distribution now extends to 1.8 M  . Both the PS20 and F12 distributions experience a stark drop in
The higher mass NSs come exclusively from stars with 𝑀CO > 7.5 frequency for BHs above 10 - 11 M  depending on the distribution,
M  . These stars happen to fall in an island of explodability in the
PS20 models where the 𝑀4 values are higher than usual. But just
having a higher 𝑀CO does not necessarily entail a higher NS mass.
Even NSs which come from progenitors with 𝑀CO greater than those
of the single star population (Fig. 3) still produce NSs with masses
again caused by the turnover in 𝑀He , still present in the binary
models, weight penalties from the IMF, and, for the F12 prescription,
the turnover in 𝑀fin for some of the binary populations (see details
in §3.1.2).
When the PS20 prescription is changed to adopt 𝑀fin as the BH,
between 1.2 and 1.6 M  . This is because of how the star responds the remnant masses shift higher (bottom panel of Fig. 7). Because the
when its envelope is entirely stripped, either by extreme wind mass stars which experience envelope stripping span a much wider range

MNRAS 000, 1–11 (2021)

8 Patton et al.
4.1 Single star distributions
PS20
F12 Fig. 8 shows both the NS and BH distributions predicted by the
PS20 and hybrid prescriptions. The two methods largely agree both
10−1
in predicted frequency and remnant mass range. The IMF weighted
average remnant mass from each is comparable, 1.39 M  for the PS20
distribution and 1.4 M  for the hybrid. The Sukhbold et al. (2016)
dM
dN

models do include fallback, but typically proto-NSs only gained a

few hundredths of solar mass.
There is also little disagreement between the two methods in the
predicted BH distribution, and the PS20 and hybrid prescriptions
predict comparable average remnant masses, 8.67 and 9.22 M  re-
10−2
4 6 8 10 12 14 16 spectively. All BHs above 9M  have nearly identical predicted fre-
Mbh
quencies, and below, discrepancies are minor. The drop off due to the
island of explodability are located at different masses, between 8.5
PS20 with Mfin
and 9 M  for hybrid, and between 5.5 - 6 M  for PS20 due to the
F12
differences in the underlying stellar models. We do not consider these
−1
10 dips absolute in their location, they are sensitive to both the adopted
stellar and supernova physics (see section 3.2 in PS20). Finally, the
lower limit on BH mass varies slightly between the two, 3.5 M  for
dM
dN

the PS20 distribution and 4 M  for the hybrid due to slight differ-
ences in the explosion landscape. Here we note that had we used the
10−2 Ertl et al. (2020) results to built the hybrid method, instead those of
Sukhbold et al. (2016), the BH distribution would have a substantial
population of mass gap remnants.

4 6 8 10 12 14 16
Mbh
4.2 Binary star distributions
Figure 7. BH distributions determined by the F12 (gold) and PS20 (blue) pre- The binary distributions for both the NSs and BHs experience the
scriptions from the binary models. The top panel shows the two prescriptions
similar discrepancies as in the single models. The hybrid NSs extend
as is whereas the bottom panel shows the two distributions if PS20 adopted
up to 1.7 M  a slightly lower mass than the PS20 models, but still
the final stellar mass as the BH mass. Adoption of 𝑀fin by PS20 fills out the
drop in frequency for BHs above 11 M and increases the maximum mass show similar behavior: clustering around a peak between 1.4 - 1.5
BH. M  . The IMF averaged NS masses remain largely the same as well,
1.38 and 1.39 M  for the PS20 and hybrid distributions respectively.
The BH distributions are largely the same too. Again, they agree
above 9 M  . Like in the PS20 and F12 distributions, there is a slight
increase in frequency for the highest mass BHs over the ones from
for binary systems than single stars, the PS20 distribution does not
the single star progenitors. This is again caused by the shallower
shift as dramatically as in the bottom panel of Fig. 5. However, the
decrease in 𝑀fin and 𝑀He in the binary models for stars above 80
remnant mass range above 11 M  increases in frequency as the extra
M  (see Fig. 1). The dip at 8.5 M  is caused by the island of ex-
mass from the envelope is added back into total BH mass, leading
plodability mentioned in §4.1. And below 8.5 M  the distributions
to higher mass BHs coming from lower initial-mass progenitors in
are comparable in their predicted frequencies but do not agree com-
addition to the stars already populating this region (top panel of
pletely. Still, the average remnant mass remains comparable, 8.16
Fig. 7). Because PS20 adopts 𝑀fin as the BH mass for all BHs, the
M  for the PS20 distribution and 8.43 M  for the hybrid.
distribution extends to a higher maximum BH mass, which comes
from a star with a lower initial mass that retains much of its envelope.
In the F12 distribution, the highest mass cores (𝑀CO > 11M  ) are
the ones which adopt 𝑀fin as the BH mass, but these stars are also
the most stripped; in most cases 𝑀fin = 𝑀He , leading to lower mass
BHs.
4 COMPARING THE PS20 AND HYBRID DISTRIBUTIONS
Because both methods predict remnant masses based on a star’s pre-
supernova core structure and are built from the results of the same
5 DISCUSSION
One of the limiting factors of comparing these distributions is that
the NS and BH mass functions are only somewhat known, making
it difficult to compare the predicted distributions to a robust set
of observations. There are predictions from theory (e.g., Pejcha &
Thompson 2015; Sukhbold et al. 2016; Kovetz et al. 2017) and
limited observations of compact object binaries (e.g., Özel et al.
2012; Özel & Freire 2016; The LIGO Scientific Collaboration et
al. 2020), which one could qualitatively compare to these or any
predicted compact object distributions, noting key similarities like
distribution shape and peak value. For example, the PS20 and hybrid
explosion models, we expect the PS20 distributions to agree much NS distributions peak near 1.4 M  and have a more normal shape
more closely to the hybrid approach than F12. Below we discuss the whereas the F12 NS distributions peak at 1.2 M  and drop off.
key differences between the single and binary NS and BH distribu- As further observations make the NS and BH mass distributions
tions. more robust, comparisons to theoretical predictions for the mass

MNRAS 000, 1–11 (2021)

 Compact Object Distribution Comparison 9

101
PS20 PS20
Hybrid 101 Hybrid

100

100
10−1
dM
dN

dM
dN
10−2
10−1

10−3

10−2 10−4
1.0 1.2 1.4 1.6 1.8 2.0 2.2 1.0 1.2 1.4 1.6 1.8 2.0 2.2
Mns, grav Mns, grav

PS20 PS20
Hybrid Hybrid
10−1 10−1
dM
dN

dM
10−2 dN 10−2

4 6 8 10 12 14 16 4 6 8 10 12 14 16
Mbh Mbh

Figure 8. The predicted distributions from single progenitors using the pre-
scriptions from PS20 (blue) and the hybrid approach (orange) for NSs (top)
and BHs (bottom). Because both methods predict remnant masses based on
the stars’ presupernova core structures, the overall agreement between the
two distributions in shape, frequency, and mass range is very good.
Figure 9. The predicted distributions from binary progenitors using the pre-
scriptions from PS20 (blue) and the hybrid approach (orange) for NSs (top)
and BHs (bottom). As with the single star distributions, the overall agreement
between the two methods’ predicted shapes, frequencies, and mass ranges are
good.

distributions will more concretely show whether any of these BPS
prescriptions produce similar distributions.
Still, we now know that a star’s explodability depends on its pre-
supernova core structure, which is not considered in the F12 pre-
scription. That is not to say that the PS20 prescription is without its
issues. Section 6 of PS20 discusses several key caveats to using bare
CO-cores. The main issue is that bare CO-cores are not perfect substi-
tutes for CO-cores embedded in stars. The temperature and pressure
gradients are steeper in bare cores than in embedded ones, which
leads to non-negligible effects on the core structure. Currently the
PS20 models are adequate for predicting bulk trends in populations
but for improved accuracy, the study would need to be redone with
CO-cores where the effects of a helium shell were also simulated.
Nevertheless, it is important to examine the implications of using
simplified prescriptions for finals fates, remnant types, and remnant
masses in BPS, especially if BPS will be used to make predictions
for observations, like merger rates from gravitational wave sources.
Some recent work is starting to explore this. Román-Garza et al.
(2020) compare the F12 delayed and rapid prescription and 1D Su-
pernova models from the N20 engine in Sukhbold et al. (2016) in
the context of progenitors of BH-NS mergers. They find that the N20
models consistently produce up to an order of magnitude more merg-
ers than F12, more consistent with the rates inferred from the first
half of the LIGO-Virgo collaboration’s third observing run (Abbott et
al. 2020a). Additionally, Ghodla et al. (2021) compares predictions
for the distributions of observed chirp masses and mass ratios for de-
tectable LIGO gravitational wave events where remnant masses are
determined by four different prescriptions, including the F12 rapid
prescription and the default BPASS prescription (see Eldridge &
Tout 2004b) in addition to two limiting cases.
More work like this is warranted, exploring which prescriptions
best reflect the observed rates of various gravitational wave sources
and the BH and NS mass functions. Recently, Schneider et al. (2021)
used the Müller et al. (2016) semi-analytic models to examine the
distributions of NSs and BHs from a suite of progenitors stripped by
binary interaction. The Müller et al. (2016) prescription is similar to
the Ertl et al. (2016, 2020) criterion adopted by the PS20 prescription
(see Sukhbold et al. 2018). However, the differences in population
explored by this work and Schneider et al. (2021) make the two
difficult to compare. Still, there is no study which compares predic-
tions for something made by every common final fate/remnant mass
prescription for the same initial population to observations of said
phenomenon. To ensure that BPS codes consistently adopt the most
accurate prescription they can, these comparisons must be explored.
There is a hierarchy of accuracy. The ideal approach for predicting
final fates or compact object distributions would be to evolve all mas-
sive stars in the simulation through to core–collapse and then try to
explode the star using a supernova explosion simulation. This is un-
feasible for now as the computational expense is quite large. Recent
studies of the single and binary progenitors of core–collapse super-
novae evolved all the way to core–collapse explore tens to hundreds
of stars (e.g., Sukhbold et al. 2016; Vartanyan et al. 2021). Doing

MNRAS 000, 1–11 (2021)

10 Patton et al.
this for millions of stars is unrealistic. The computational expense of
end-stage modelling could be mitigated in part by probing core struc-
ture at core collapse and making predictions for final fate/remnant
mass using parameters calibrated from 1D core–collapse supernovae
models, as in the two-parameter criterion from Ertl et al. (2016,
2020).
But BPS codes only evolve stars to central carbon ignition (or
slightly later or slightly earlier). The next best way to proceed would
be to back out a presupernova core-structure using the PS20 models.
Then the user could adopt their preferred method for determining
final fate/remnant mass based on the core structure. However, since
most BPS codes do not actively evolve their stars, they cannot use the
PS20 models. Instead, we recommend adopting the hybrid prescrip-
tion in rapid BPS codes because it can be implemented easily, while
still incorporating the strong dependence final fate has on presuper-
nova core structure. The user can also pick their preferred explosion
models to construct the hybrid matrix.
6 CONCLUSIONS
We present a comparison of the compact object distributions pre-
dicted by three separate remnant prescriptions, all applied to the
same populations of solar metallicity single and binary stars mod-
eled by BPASS. The F12 prescription depends solely on 𝑀CO and
the final mass of the star whenever a star’s evolution terminates.
For BPS codes, which commonly employ it, that cutoff is at or near
central carbon ignition. The PS20 prescription approximates a pre-
supernova structure by interpolating over a grid of CO-cores evolved
from central carbon ignition to core–collapse, using values of 𝑀CO
and 𝑋C taken from the BPS models at carbon ignition. By applying
a core structure-based explosion criterion onto the final core, in this
case the Ertl et al. (2016, 2020) two-parameter explosion criterion,
we predict final fates and remnant masses based on the presuper-
nova core structure. Finally, we adopt a hybrid approach which uses
𝑀He measured at the evolutionary cutoff to interpolate over a table
of presupernova 𝑀He and associated remnant masses from detailed
evolution and explosion models.
The differences between the F12 and PS20 prescriptions, outlined
in §3.1 and §3.2, are stark. The F12 prescription predicts a near
continuous distribution of remnant masses, including high mass NSs
agreement between two different methods which both rely, either ex-
plicitly (PS20) or implicitly (hybrid), on presupernova core structure
points to the necessity of considering that structure when determin-
ing how a star dies and the properties of the remnant it leaves behind.
There is a strong correlation between final core structure and final
fate. There is not a strong correlation between 𝑀CO and/or 𝑀fin and
final fate.
Qualitatively, the shape and peak of the PS20 and hybrid distri-
butions agree better with other theoretical predictions and limited
observations of the NS and BH mass function, compared to F12, the
most commonly utilized prescription in BPS studies. Given the in-
tegral role played by the presupernova core structure in determining
final fates of massive stars and their associated remnant properties,
we strongly encourage future BPS calculations to consider and in-
clude this structural dependence and utilize modern neutrino-driven
explosion results for an improved accuracy of their predictions. As
we have shown, the PS20 and hybrid methods are two possible ways
to do this, depending on the type of BPS code and the purpose of the
study.
7 ACKNOWLEDGEMENTS
The authors would like to thank E. Stanway and A. Chrimes for their
contributions to the BPASS models. JJE acknowledge support from
the University of Auckland and also the Royal Society Te Apārangi of
New Zealand under the Marsden Fund. Support for this work comes
from NASA grant 80NSSC19K0597 awarded to M. Pinsonneault.
TS was supported by NASA through the NASA Hubble Fellowship
grant #60065868 awarded by the Space Telescope Science Institute,
which is operated by the Association of Universities for Research in
Astronomy, Inc., for NASA, under contract NAS5-26555.
8 DATA ACCESSIBILITY
All models used in the PS20 prescription (KEPLER models, the MESA
inlist, and the complete tables of core diagnostics) are available for
download at http://doi.org/10.5281/zenodo.3839747. The
specific BPASS data files containing core masses and composition
information for this study, as well as a list of the missing models,
(𝑀grav ≈ 2.2 M  ) and high frequency of low mass (𝑀bh ≈ 2.5 M  )
BHs (see Fig. 3, Fig. 5, Fig. 6, and Fig. 7). The PS20 prescription
shows that core structure limits the possible NS masses and that low-
mass, mass gap black holes are much less frequent. NSs produced
from higher mass CO-cores, which F12 does not predict, are the
direct result of the increase in 𝑋C caused by envelope stripping as
well as islands of explodability, neither of which F12 is sensitive to.
While the BH mass range covered by the two prescriptions’ distri-
butions is roughly identical, the distributions themselves differ quite
are available upon request. Standard BPASS output is available at
https://bpass.auckland.ac.nz/index.html.
REFERENCES
Abbott, R., Abbott, T. D., Abraham, S., et al. 2020, arXiv:2010.14527
De Marco, O. & Izzard, R. G. 2017, Publ. Astron. Soc. Australia, 34, e001.
doi:10.1017/pasa.2016.52
Eggleton, P. P. 1971, MNRAS, 151, 351. doi:10.1093/mnras/151.3.351
dramatically around BH masses of 10-12.5 M  . This is the direct
result of envelope stripping driving down the available mass of the
star to form a remnant (see Fig. 2 and Fig. 3). The PS20 models do
not include any mass from the envelope in their BH mass calculation,
causing no frequency drop in this range. The frequency drop in the
PS20 distributions at slightly higher BH masses is due to penalties
from the IMF and a small turnover in 𝑀He caused by extreme mass
loss.
The differences between the PS20 and hybrid distributions are
minimal and largely due to differences in the model physics adopted
by PS20 and Sukhbold et al. (2016), from where we draw our models
to construct the hybrid matrix (see Fig. 8 and Fig. 9). This strong
MNRAS 000, 1–11 (2021)
Eldridge, J. J., Izzard, R. G., & Tout, C. A. 2008, MNRAS, 384, 1109.
doi:10.1111/j.1365-2966.2007.12738.x
Eldridge, J. J. & Stanway, E. R. 2009, MNRAS, 400, 1019.
doi:10.1111/j.1365-2966.2009.15514.x
Eldridge, J. J., Stanway, E. R., Xiao, L., et al. 2017, Publ. Astron. Soc.
Australia, 34, e058. doi:10.1017/pasa.2017.51
Eldridge, J. J. & Tout, C. A. 2004, MNRAS, 353, 87
Ertl, T., Janka, H.-T., Woosley, S. E., et al. 2016, ApJ, 818, 124.
doi:10.3847/0004-637X/818/2/124
Ertl, T., Woosley, S. E., Sukhbold, T., et al. 2020, ApJ, 890, 51.
doi:10.3847/1538-4357/ab6458
Fryer, C. L., Belczynski, K., Wiktorowicz, G., et al. 2012, ApJ, 749, 91.
doi:10.1088/0004-637X/749/1/91
 Compact Object Distribution Comparison 11
Ghodla, S., van Zeist, W. G. J., Eldridge, J. J., et al. 2021, arXiv:2105.05783
Hurley, J. R., Pols, O. R., & Tout, C. A. 2000, MNRAS, 315, 543.
doi:10.1046/j.1365-8711.2000.03426.x
Hurley, J. R., Tout, C. A., & Pols, O. R. 2002, MNRAS, 329, 897.
doi:10.1046/j.1365-8711.2002.05038.x
Jayasinghe, T., Stanek, K. Z., Thompson, T. A., et al. 2021, arXiv:2101.02212
Kovetz, E. D., Cholis, I., Breysse, P. C., et al. 2017, Phys. Rev. D, 95, 103010.
doi:10.1103/PhysRevD.95.103010
Kroupa, P. 2001, MNRAS, 322, 231. doi:10.1046/j.1365-8711.2001.04022.x
Laplace, E., Justham, S., Renzo, M., et al. 2021, arXiv:2102.05036
Lattimer, J. M. & Prakash, M. 2001, ApJ, 550, 426. doi:10.1086/319702
Lattimer, J. M. & Yahil, A. 1989, ApJ, 340, 426. doi:10.1086/167404
The LIGO Scientific Collaboration, the Virgo Collaboration, Abbott, R., et
al. 2020, arXiv:2010.14533
Lovegrove, E. & Woosley, S. E. 2013, ApJ, 769, 109. doi:10.1088/0004-
637X/769/2/109
Moe, M. & Di Stefano, R. 2017, ApJS, 230, 15
Müller, B., Heger, A., Liptai, D., et al. 2016, MNRAS, 460, 742
Nugis, T. & Lamers, H. J. G. L. M. 2000, A&A, 360, 227
O’Connor, E., & Ott, C. D. 2011, ApJ, 730, 70
Özel, F. & Freire, P. 2016, ARA&A, 54, 401. doi:10.1146/annurev-astro-
081915-023322
Özel, F., Psaltis, D., Narayan, R., et al. 2012, ApJ, 757, 55. doi:10.1088/0004-
637X/757/1/55
Patton, R. A. & Sukhbold, T. 2020, MNRAS, 499, 2803.
doi:10.1093/mnras/staa3029
Pejcha, O. & Thompson, T. A. 2015, ApJ, 801, 90. doi:10.1088/0004-
637X/801/2/90
Pols, O. R., Tout, C. A., Eggleton, P. P., et al. 1995, MNRAS, 274, 964.
doi:10.1093/mnras/274.3.964
Román-Garza, J., Bavera, S. S., Fragos, T., et al. 2020, arXiv:2012.02274
Salpeter, E. E. 1955, ApJ, 121, 161
Sana, H., de Mink, S. E., de Koter, A., et al. 2012, Science, 337, 444
Sana, H., de Koter, A., de Mink, S. E., et al. 2013, A&A, 550, A107
Schneider, F. R. N., Podsiadlowski, P., & Müller, B. 2021, A&A, 645, A5.
doi:10.1051/0004-6361/202039219
Sravan, N., Marchant, P., & Kalogera, V. 2019, ApJ, 885, 130.
doi:10.3847/1538-4357/ab4ad7
Stanway, E. R. & Eldridge, J. J. 2018, MNRAS, 479, 75.
doi:10.1093/mnras/sty1353
Sukhbold, T., & Adams, S. 2020, MNRAS, 492, 2578
Sukhbold, T., Ertl, T., Woosley, S. E., et al. 2016, ApJ, 821, 38
Sukhbold, T. & Woosley, S. E. 2014, ApJ, 783, 10. doi:10.1088/0004-
637X/783/1/10
Sukhbold, T., Woosley, S. E., & Heger, A. 2018, ApJ, 860, 93
Thompson, T. A., Kochanek, C. S., Stanek, K. Z., et al. 2019, Science, 366,
637. doi:10.1126/science.aau4005
Timmes, F. X., Woosley, S. E., & Weaver, T. A. 1996, ApJ, 457, 834.
doi:10.1086/176778
Ugliano, M., Janka, H.-T., Marek, A., & Arcones, A. 2012, ApJ, 757, 69
Vartanyan, D., Laplace, E., Renzo, M., et al. 2021, arXiv:2104.03317
Woosley, S. E. 2019, ApJ, 878, 49. doi:10.3847/1538-4357/ab1b41
Woosley, S. E. & Heger, A. 2015, ApJ, 810, 34. doi:10.1088/0004-
637X/810/1/34
Woosley, S. E., Heger, A., & Weaver, T. A. 2002, Reviews of Modern Physics,
74, 1015
Woosley, S. E., Sukhbold, T., & Janka, H.-T. 2020, ApJ, 896, 56.
doi:10.3847/1538-4357/ab8cc1
Zapartas, E., de Mink, S. E., Justham, S., et al. 2019, A&A, 631, A5.
doi:10.1051/0004-6361/201935854

This paper has been typeset from a TEX/LATEX file prepared by the author.

MNRAS 000, 1–11 (2021)